The Cato Institute published a paper yesterday entitled “Effective Counterterrorism and the Limited Role of Predictive Data Mining” by Cato’s Jim Harper (whose book Identity Crisis I reviewed here last month) and fellow IBMer/blogger Jeff Jonas. The paper argues against the value of pattern-based predictive data mining as a tool for counterterrorism, instead favoring tools that investigate links among and outward from known suspects, using intelligence and detection tools within a more robust information-sharing environment.
This seems to be an emerging consensus viewpoint about predictive data mining; indeed, as the National Journal reported recently, a recent procurement document admitted the challenges that the US government has faced in developing reliable predictive models for counterterrorism applications. There is still a value in using predictiving modeling in limited security applications – e.g. looking for trends among cleared government personnel that would suggest counterintelligence activity (as mentioned in this post) – but in general, other forms of data analysis are more suited for counterterrorism.
Overall, a solid piece. For more on this topic, I would again recommend Mary DeRosa’s Data Mining and Data Analysis for Counterterrorism.